Methods and apparatus to train interdependent autonomous machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for training autonomous machines in industrial settings are complex and limited in scalability due to the need for independent training of each robot, which fails to account for interdependencies between robots, leading to increased likelihood of system failures and high training complexity.
Innovation Solution
Implementing deep reinforced learning across robot controllers to enable self-training and account for interdependencies between robots, using machine learning engines like convolutional neural networks to adapt coefficients and learn from interactions, reducing the need for human intervention and simplifying training processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If independent training of each robot is used, then training process is simple for individual robots, but system reliability deteriorates due to increased likelihood of failures and inability to account for interdependencies
Solution Approach 1:
The patent merges the training processes of multiple robots into a unified deep reinforced learning system where robot controllers are trained together to account for interdependencies. The machine learning engine processes sensor data from multiple robots simultaneously and generates coordinated control actions, transforming independent training into a collaborative training framework that maintains simplicity while improving reliability.
Solution Approach 2:
The patent implements a universal machine learning engine that serves multiple functions: it processes sensor data from any robot in the group, generates control actions for multiple robots, and adapts to different collaborative tasks. This multi-functional approach allows the same training framework to handle various robot configurations and tasks, maintaining ease of manufacture while ensuring system reliability through coordinated learning.
2Reliability
If traditional training methods are used to accommodate interdependencies, then system reliability improves, but training complexity far exceeds order N
Solution Approach 1:
The patent implements self-service training where the robot controllers automatically learn interdependencies through deep reinforced learning without requiring complex manual programming or extensive human intervention. The machine learning engine autonomously processes sensor data, identifies interdependencies, and adjusts control policies, achieving high reliability with training complexity that scales linearly rather than exponentially.
Solution Approach 2:
The patent uses continuous feedback loops where sensor data from robot interactions during collaborative tasks is fed back to the machine learning engine. This feedback mechanism allows the system to automatically learn and adapt to interdependencies in real-time, achieving high reliability through data-driven learning rather than complex pre-programming, keeping training complexity manageable.
3Productivity
If more robots are added to robotic cells, then productivity increases, but training complexity and difficulty of scaling increase
Solution Approach 1:
The patent creates a universal training framework that can handle any number of robots through the multi-functional machine learning engine. The same deep reinforced learning architecture processes sensor data from N robots, generates coordinated control actions, and adapts to different task configurations, enabling scalable productivity improvement without proportional increase in training complexity.
Solution Approach 2:
The patent achieves scalability by changing the parameter N (number of robots) without fundamentally altering the training framework. The deep reinforced learning system naturally adapts to different group sizes through its flexible architecture, allowing productivity to scale with robot count while training complexity remains manageable through the unified approach.
Data Source
AI summary
Methods and apparatus to train interdependent autonomous machines are disclosed. An example method includes performing an action of a first sub-task of a collaborative task with a first collaborative robot in a robotic cell while a second collaborative robot operates in the robotic cell according to a first recorded action of the second collaborative robot, the first recorded action of the second collaborative robot recorded while a second robot controller associated with the second collaborative robot is trained to control the second collaborative robot to perform a second sub-task of the collaborative task, and training a first robot controller associated with the first collaborative robot based at least on a sensing of an interaction of the first collaborative robot with the second collaborative robot while the action of the first sub-task is performed by the first collaborative robot and the second collaborative robot operates according to the first recorded action.